Brandon Witham

I'm not seeing any obvious ways in which this can be sped up. Made a small correction to the call to `AreaCriterion`. Will leave open for now in case any...

#3441 adds some address support to building conflation...it could maybe still be done better if implemented at the training phase instead.

~~I believe, but am not certain, that rubber sheeting always takes into account all data loaded in the input datasets. It doesn't matter whether you turn any of the conflation...

Yeah, from the statements I crossed out its probably not clear. I believe your initial statement is right (but still need to verify) that the more features included in general,...

Furthermore, I think the denser the input data, the more the higher you would want to make your min tie point requirement.

see #3534 and #3530

Short of reworking the algorithm itself, which is unlikely at this time, I can only think of looking into caching and maybe some basic efficiency checks for this. **Maybe** there's...

Not really seeing much that can be optimized here. `EdgeMatchSetFinder::_addEdgeMatches` specifically is where most of the time is spent. Since its a recursive algorithm not seeing much that can be...

Found this setting: `network.edge.match.set.finder.max.iterations`, which controls the recursion in `EdgeMatchSetFinder::_addEdgeMatches`. It doesn't look like an optimal value for it has ever been found with `optimize-network-conf`, so I'll do that and...

Dropping `network.edge.match.set.finder.max.iterations` drastically from 20 to 1, reduces the runtime for gaal large by ~300%. This of course, says nothing for what effect it has on the conflation output quality...so...